How 168EP generates esports match picks · Algorithm V1 · Last updated November 2026
168EP uses a statistical model that combines three independently sourced data inputs to estimate each team's probability of winning a match. When the model's confidence in one team is sufficiently higher than the implied market probability, a pick is issued.
All picks are logged with full data snapshots and compared against actual results. Past performance is publicly visible on the Results page.
Odds from VSGG are converted to implied probability by removing the bookmaker margin:
This gives a baseline probability that reflects the consensus view of professional bookmakers.
Three factors are combined to produce a model score for each team:
| Factor | Weight | Data Source | Description |
|---|---|---|---|
| Market Odds | 60% | VSGG | Bookmaker implied probability after margin removal |
| Recent Form | 30% | Esports8 | Weighted win rate across last 5 matches (most recent weighted highest) |
| Head-to-Head | 10% | Esports8 | Historical win rate between the two specific teams |
When multiple odds snapshots exist for the same match, the algorithm detects movement:
A pick is issued only when the model score gap exceeds a minimum threshold:
Confidence is scaled proportionally to the gap size, capped at 100%.
168EP issues NO PICK when any of the following conditions apply:
Data is refreshed every 6 hours. Odds snapshots are saved on each collection to enable movement tracking.
Games covered: CS2 · Dota 2 · Valorant · League of Legends
Every pick is stored with:
Actual results are collected automatically after each match. View the full pick history →